Frequently Asked Questions

General

What is GPUniq?

GPUniq is a GPU Meta-Cloud platform that aggregates compute capacity from multiple providers. Rent GPUs through three modes: Marketplace, Dex-Cloud, and Burst. Access persistent storage, LLM models, and manage everything through a Python SDK or REST API.

How is GPUniq different from AWS/GCP/Azure?

• Lower prices: Up to 70% cheaper than major cloud providers
• More GPU options: Consumer and professional GPUs from thousands of providers
• Flexible billing: Hourly, daily, weekly, or monthly pricing
• No commitments: Rent for as long as you need, stop anytime
• Python SDK: pip install GPUniq for full programmatic control

What GPUs are available?

The marketplace includes 5000+ servers with GPUs including:

• Consumer: RTX 3090, RTX 4090, RTX 5090
• Professional: A100, H100, A6000, L40S
• Availability varies in real-time

Authentication

How do I authenticate?

Two methods:

• API Key (recommended): Create in dashboard under LLM API Keys. Use X-API-Key header or Python SDK.
• JWT Token: For web dashboard, uses Authorization: Bearer header.

What's the rate limit?

120 requests/minute per API key. The Python SDK automatically retries on rate limit.

GPU Compute

What are the three deployment modes?

Can I use Docker images?

Yes. Specify a docker_image when creating orders (e.g., pytorch/pytorch:latest).

How do I connect to my GPU?

Use SSH with credentials from your dashboard:

ssh root@{host} -p {port}

Or manage SSH keys via the Settings page for passwordless access.

Payments

What payment methods are supported?

• USDT (TRC20): Crypto deposits via Tron network
• Stripe: International bank cards
• YooKassa: Russian bank cards and local methods

How does billing work?

Balance is deducted in real-time as you use resources. Longer pricing commitments (day/week/month) offer 20-40% savings over hourly rates.

Can I withdraw my balance?

Yes, withdraw in USDT to any TRC20 wallet address.

Storage

How do I persist data between instances?

Use Volumes — persistent storage that survives instance stop/start/delete. Create a volume, upload data, and attach it to any instance.

What's the maximum upload size?

500 MB per file via API. For larger files, use SCP/rsync directly to the instance.

LLM API

What models are available?

Models from OpenAI, Qwen, DeepSeek, Meta Llama, Mistral, NVIDIA, and more. See LLM API docs or call client.llm.models().

How do I use the LLM API?

from gpuniq import GPUniq
client = GPUniq(api_key="gpuniq_your_key")
response = client.llm.chat("openai/gpt-oss-120b", "Hello!")

Technical

Do GPUs come with CUDA installed?

Most providers have CUDA pre-installed. Verify with nvidia-smi after connecting.

Is my data secure?

Each rental runs in an isolated Docker container. When you delete your instance, data is removed. Use Volumes for data you need to keep.